Individual, household and gender preferences for social transfers

Available online at www.sciencedirect.com

Journal of Economic Psychology 29 (2008) 348–359

www.elsevier.com/locate/joep

Individual, household and gender preferencesfor social transfers

Liam Delaney a, Francis O’Toole b,*

a Geary Institute, University College Dublin, Irelandb Department of Economics, Trinity College Dublin, 2 College Green, Dublin, Ireland

Received 15 February 2007; received in revised form 3 September 2007; accepted 7 September 2007Available online 19 September 2007

Abstract

This paper reports the results of a nationally representative survey that explored willingness topay extra taxes for increased levels of social transfers in Ireland. Respondents differ in their interpre-tation of willingness-to-pay questions with respect to the relevant income constraint being the indi-vidual’s budget or the individual’s household budget, with an important explanatory variable in thisinterpretation being the level of financial management integration within the household. Respon-dents are also shown to take intra-household bargaining considerations into account when statingpreferences for specific redistributive policies; a significant gender difference emerges for a specificfiscal policy (child benefit) which alters the intra-household entitlement to income.� 2007 Elsevier B.V. All rights reserved.

JEL classification: Z11; C42

PsycINFO classification: 2260; 3040

Keywords: Survey methods; Household economics

1. Introduction

The standard economic theoretical models of voter preferences assume that only thosewho gain monetarily either directly or indirectly (e.g. via their children) from different

0167-4870/$ - see front matter � 2007 Elsevier B.V. All rights reserved.

doi:10.1016/j.joep.2007.09.003

* Corresponding author. Tel.: +353 1 896 1068.E-mail address: [email protected] (F. O’Toole).

mailto:[email protected]

L. Delaney, F. O’Toole / Journal of Economic Psychology 29 (2008) 348–359 349

welfare schemes will support re-distributive policies, although there exists a literature (e.g.Fong, 2001; Hochman & Rodgers, 1969) which incorporates more flexible forms on utilitywith respect to the consumption and/or utility of others. For example, in the Hochmanand Rodgers (1969) formulation, individuals have (non-kin) altruistic preferences butdirect transfers of income by private citizens are a socially inefficient mechanism forachieving the necessary redistribution, which motivates a role for government in providingredistribution through a progressive income tax.

In addition, the vast majority of empirical papers on stated preferences for inequalityhave been conducted on large international data sets that can only analyse general atti-tudes to redistribution (e.g. Scheve & Slaughter, 2006) rather than preferences for, andwillingness to pay (WTP) for, more specific policies of redistribution and social transfers.

In contrast to these standard literatures, this paper reports on a nationally representa-tive survey in the Republic of Ireland (‘‘Ireland’’) that attempted to assess stated prefer-ences, and in particular WTP, for specific social transfers.1 Once WTP for specificproposals is being considered, however, it is necessary to develop a more contextualisedmodel of the manner in which individuals, as parts of households, make choices, particu-larly if the goal of the analysis is to construct aggregated measures of valuation. Therefore,the first task of this paper is, in the context of the general literature on contingent valua-tion methodology (CVM) and building on earlier work in Delaney and O’Toole (2006), toexplore the issue of eliciting preferences from individuals, bearing in mind that most indi-viduals are parts of households. The influence of gender and the various possible house-hold financial management systems are considered in this regard in this paper.

Tests of the pooled income hypothesis have consistently shown that differences inincome between husbands and wives have significant effects on the allocation of incomeacross different expenditure items and, in particular, on household-related outcomes suchas child health (e.g., Lampietti, 1999). This paper specifically tests whether or not wivesfrom high income households are more opposed than husbands from high-income house-holds to a set of policies that would make child benefit, a transfer generally paid to themother, more progressive in household income. Therefore, the second task of this paperis to provide a unique test of the pooled income hypothesis.

The remainder of this paper is structured as follows. Section 2 reviews the relevantexisting literature. Section 3 describes the structure and administration of the nationwidesurvey designed to examine general and specific preferences for redistribution. Section 4examines the results while Section 5 offers concluding comments.

2. Individual and household willingness to pay for social transfers

2.1. Willingness to pay: Respondents’ self-perceived agency

There are several methods of eliciting household WTP, and a full taxonomy of differentelicitation methods should involve interacting the traditional taxonomy with permutationsof the different manners (e.g. order) in which the members of a household could conceiv-ably be interviewed. Delaney and O’Toole (2006) analysed the issue of eliciting individual

1 For an overview of contingent valuation in the more general non-market or public goods setting, see Mitchelland Carson (1989) or Champ, Boyle, and Brown (2003).

350 L. Delaney, F. O’Toole / Journal of Economic Psychology 29 (2008) 348–359

and household willingness to pay from individual respondents in a standard contingentvaluation setting. Following Quiggin (1998), the authors argued that this issue had crucialsignificance for the contingent valuation literature and that many studies were rendered atthe least very imprecise by failing to adequately address the issue. The authors showed thatrespondents varied significantly in the way they modelled a standard WTP question withsome respondents choosing to provide individual WTP and some respondents choosing toprovide (or estimate) household WTP. In the specific context of public service broadcast-ing, Delaney and O’Toole (2004) showed that this individual WTP versus household WTPdecision was influenced significantly by socio-demographic characteristics such as gender,age and the presence or absence of children in the household.

In this paper, the concept of respondents’ self-perceived agency in valuation studies isfurther developed and empirically tested. The key issue is the extent to which the respon-dent to a WTP question views the relevant budget constraint as being her/his own share ofthe household budget or the total household budget. In particular, there may be significantheterogeneity (e.g. with respect to gender) in perceived entitlement to spend from thehousehold budget. Strand (2004) addresses the related adding-up (of individual’s willing-ness to pay for household public goods) issue from a theoretical perspective.

More specifically, two related questions are addressed in the context of the standardopen-ended WTP question. First, do individual respondents process the standard open-ended question, ‘‘What is your maximum WTP?’’ as being her/his own individual WTPfrom her/his own budget constraint (however constructed or visualised) or as being her/his (estimate of) household WTP from the household budget constraint? Second, do indi-vidual respondents process the open-ended question, ‘‘What is your maximum household

WTP?’’ as being her/his (estimate of) household WTP from the household budget con-straint or, and notwithstanding the explicit reference to the word ‘‘household’’ in the ques-tion, as being her/his own individual WTP from her/his own budget constraint (again,however constructed or visualised)?

In effect, Delaney and O’Toole (2006) only analysed how respondents modelled the firstof these two questions. The basic willingness to pay scenario, which centred on servicesprovided by the Irish public service broadcaster (RTE) was presented as follows:

Q.13 ‘‘Thinking of a situation where there was no licence fee and you had a choice ofeither paying to receive RTE’s services or not paying and not receiving RTE’s ser-vices. Bearing in mind that any money that you spend is money that you could spendon other goods and services, what would be the maximum amount of money youwould be prepared to pay each month in order to receive RTE’s services? (Do notprompt)’’

If appropriate, the respondent was then asked the follow-up question(s).

Q.14 (Only ask this question if there is more than one person in the respondent’shousehold) ‘‘Which of the following best describes your answer to Q.13? (Tick one)

• This is the most you personally would be willing to pay. (If yes, then ask Q.15).• This is the most your entire household would be willing to pay. (If yes, then skip Q.15).

Q.15 In light of your answer to Q.14, what do you think is the maximum amount ofmoney your household would be willing to pay each month to receive RTE’s services?’’


The aim of that analysis was to demonstrate the aggregation biases that result whenindividual respondents are left, to at least some extent, to model their own agency. How-ever, the aim of this paper is to demonstrate that even when the researcher more explicitlyframes the agency of the respondent (i.e. your household), the respondent may not, per-haps for a variety of reasons, model the task set in the manner dictated by the questionset-up. For example, it would seem possible that individuals as part of households maysuppress (from consciousness) income-sharing rules within the household for psychologi-cal or cultural reasons.

The determinants of how individual respondents model the standard WTP questionscan be usefully analyzed by estimating conditional distributions on demographic factorssuch as gender, age and occupation. In addition, the importance of financial managementintegration within the household is a measure that is highlighted and explored in intra-household allocation empirical analysis (e.g. Burgoyne, 1995; Pahl, 1995). For example,respondents who are in a financially integrated relationship may be considered less likelyto distinguish between her/his own ‘‘individual’’ WTP and ‘‘household’’ WTP, whereasthose in a relationship that is not financially integrated may be considered more likelyto distinguish between her/his ‘‘individual’’ WTP and her/his ‘‘household’’ WTP. Kirchler,Rodler, Holzl, and Meier (2001) provide an excellent introduction to, and overview of, theliterature on decision-making within close relationships. In particular, the authors providean excellent review of the over 20 years of research that has been conducted in this area inthe Universities of Linz and Viennna, Austria (e.g. Vienna Diary Study, 1996–1999);Chapter 2 on Love, Money and Everyday Matters provides a very useful backdrop for thispaper.

2.2. Intra-household factors and preferences for transfers

Tests of the pooling hypothesis have consistently demonstrated a difference in the effectof income controlled by either husbands or wives on different outcomes, such as childhealth and nutrition (Schultz, 1990; Thomas, 1990) and expenditure on alcohol andtobacco (Hoddinott & Haddad, 1995; Phipps & Burton, 1992). Lampietti (1999) uses dif-ferences between husbands and wives in WTP for malarial prevention to test the ‘‘pooledincome hypothesis’’.2 At a general level, several authors have assessed the concept of a‘‘gender gap’’ in fiscal preferences with the evidence apparently pointing against a markedgender difference. A number of recent papers have also examined gender differences inpreferences for environmental goods (e.g. Dupont, 2004). Examining such general prefer-ences, while interesting, fails to disentangle from an economic perspective the reasons whyone would expect gender differences. To do this, one should examine specific schemes andtheir implication for the distribution of entitlements within the household.3

One appropriate and potentially very interesting test of the shared income hypothesis inthe Irish context is preferences for the conditionality of child-benefit payments. Child-ben-efit payments in Ireland are universal and are generally paid to the mother. One of the chiefarguments against making child-benefit conditional on (individual or household) income is

2 Co-operative and non-cooperative bargaining models of household behaviour and preferences offeralternatives to the pooled income hypothesis, and can also be employed to explore preferences for wide-scalegovernment activity.

3 In this regard, see also Alvarez and McCaffrey (2003).


that given that it is a payment to the mother, reducing it to any segment in society wouldhave negative consequences for the mother in the intra-household allocation process.4

The pooled income hypothesis, whereby households maximize a common-utility func-tion, puts testable restrictions on parameters describing individual preference structures onthe survey data. If preference structures are formed in this type of world, then there is noreason to believe that men and women, holding income constant, would have differentpreferences for the proposal to make child benefit payments partly dependent on house-hold income. However, a model where degree of entitlement to income within a householddetermines intra-household allocations would predict an interaction between income andsupport for conditionality. Specifically, women from higher income households, being thegroup that would lose effective entitlement to income, should be more opposed to the pro-posal that child-benefit be made conditional on household income than men from higherincome households.

3. Survey design and scenario

3.1. Administration

The specific questions examined in this paper form part of a survey eliciting preferencesfor social welfare expenditures in Ireland. The nationally representative survey, of thoseaged 15+, was based on 1159 face-to-face individual respondent interviews (in the respon-dents’ homes) carried out by Lansdowne Market Research in June 2004. There was quotacontrolling based on age, gender, place of residence and occupation; however, the charac-teristics of non-respondents were not recorded. The interviewer implemented (randomly)the split-sample procedure (‘‘you’’ versus ‘‘your household’’ – see below for furtherdetails). The questionnaire used in the nationwide survey consisted of 25 questions.5

The questions were ordered such that respondents were initially asked to consider generalattitudes to government spending (on Social and Family Affairs, Health and Education)and taxation, and were then asked the WTP question(s). More detailed questions on spe-cific social welfare schemes and, in particular, child benefit, unemployment assistance andold age pensions, were then asked.

3.2. Willingness to pay scenarios and level of financial integration

The WTP question has the disadvantage that it does not specify a precise vector of ben-efits deriving from the extra (if any) tax. Pilot testing indicated that respondents inter-preted the question as demanding a monetary amount to achieve improvements tosocial welfare benefits. Importantly, informal pilot-tests demonstrated considerable lackof clarity among respondents as to whether the amount suggested come from their ownbudget or their household budget and that differential phrasing of the initial question

4 Indeed, proposals to tax child benefit in Ireland are met with considerable opposition. For example, inresponse to a 2002 proposal, the National Women’s Council of Ireland responded, ’’The National Women’s

Council of Ireland (NWCI) is astonished and appalled that the Government is considering taxing Child Benefit . . .

We will not stand by and watch the Government scapegoat children and women’’. (NWCI Press Release, Monday 16September 2002).

5 The questionnaire is available from the authors upon request.


did not solve this problem. After several clarifications, pilot-test respondents settled on avaluation that was frequently far different from the initial amount. The practicalities of theinterview process in the nationwide sample precluded the use of detailed interview proto-cols necessary to more fully explore all the interesting issues raised.

A 2 · 2 split-sampling procedure was employed. The scenario went as follows, with halfof the respondents being asked Version A and half of the respondents being asked Version B:

Version A: Q5a Would you be willing to pay more money in taxes each week in order tosupport extra social welfare spending on children, the disabled, pensioners, carers, theunemployed and those on low incomes? In other words, imagine that the government pro-posed increasing spending on these schemes and paying for these increases by increasingtaxes on products and services and this were to cost you money. How much extra money,at a maximum, would you be willing to pay per week?

Version B: Q5a Would your household be willing to pay more money in taxes each weekin order to support extra social welfare spending on children, the disabled, pensioners,carers, the unemployed and those on low incomes? In other words, imagine that the gov-ernment proposed increasing spending on these schemes and paying for these increasesby increasing taxes on products and services and this were to cost your household money.How much extra money, at a maximum, would your household be willing to pay per week?

Both groups of respondents were then asked the follow-up question:

6 Pawage wmoneythe huor neaincom

Q5b Which of the following best represents your response to the above question?

(i) This is the total amount of extra money that my household would be willing to pay.(ii) This is the total amount of extra money that I would be willing to pay.

Respondents who answered (ii) to Q5b were asked the following question,

Q5c How much money at a maximum would your household be willing to pay?

Standard socio-demographic variables such as gender, age, income, and education wereascertained. There was also a question that asked respondents who were married or livingas married to state the manner in which their household conducted its finances.6

Q17 Which of the following statements best describes how you and your partnerconduct your financial affairs?

(i) We have completely separate finances that we rarely discuss.(ii) We have separate accounts.

(iii) We have joint accounts.(iv) We have joint accounts and conduct all our finances together.(v) We have joint accounts but what we do with the rest of our money, we decide as

individuals.

hl (1995, p. 366) identifies a number of different income allocation strategies among couples: female whole

here the women controls the allocation of the total wage (perhaps minus the male’s personal spending); male whole wage where the male controls the allocation of the total wage; housekeeping allowance wheresband gives the wife a fixed sum to manage housekeeping activities; income pooling, where there is completer-complete income sharing; and, independent management systems where both partners have their owne and finances are conducted separately.


3.3. Preferences for government expenditures and attitudes towards conditionality

Specific questions analysed in this paper concern the respondent’s preferences for thescale of general social welfare expenditures and the scale of specific child benefit payments.Respondents were asked (Q2) their preferences, on a seven-point scale (‘‘Take away a lot,. . . , Add a lot) for increased or decreased expenditure on each of the following categoriesof government expenditure: social welfare, education and health. Respondents were alsoasked (Q6) their preferences, again on a seven-point scale, for increased or decreasedexpenditure on a large number of specific social welfare schemes, e.g. child benefit, old-age pension and carer’s allowance.

In order to explore specifically the pooled income hypothesis, attitudes toward the con-ditionality, or otherwise, of child-benefit payments are explored in particular detail.Respondents were asked (Q10) to rate on a seven-point scale their level of agreement witha series of statements including:

‘‘Child benefit should only be paid to those who need it’’.‘‘Child benefit should be paid to everyone with children regardless of household income’’.‘‘Child benefit should be part of taxable income’’; and,‘‘People with high incomes should be given less child benefit than people with low

incomes’’.

4. Results

4.1. Individual/household willingness to pay

Of the 1159 respondents who were asked the WTP question, 129 (11.1%) did notrespond (11.1%). There was no statistically significant difference between the two versionsin this regard (12% versus 10.3%, p > 0.36). The majority of respondents to both versionsof the question responded that they would not be willing to pay any amount of extra tax-ation to finance increases in social welfare expenditure. In Version 1, where respondentswere first asked to give ‘‘your WTP’’, 339 of 507 respondents (66.9%) responded that theywould not be willing to pay extra taxation to fund increases in social welfare provision,while in Version 2, where respondents were asked to give ‘‘your household’’ WTP, 353of 523 (67.5%) of respondents responded that they would not be willing to pay extra tax-ation to fund increases in social welfare provision. Again, there was no statistically signif-icant difference between the two versions in this regard.

The responses to the follow-up question, asking respondents whether their amount rep-resented individual or household WTP, are displayed in Table 1. This question wasrestricted to the respondents (a total of 876) who had previously indicated that they weremarried or living as married.7 In total, 70.3% of respondents who answered either version

7 Respondents were not asked to distinguish between ‘‘married’’ and ‘‘living as married’’; we thank oneanonymous reviewer for raising this important issue. Ideally, it would have been very useful to have suchinformation for present purposes. However, it should be noted that market research companies in Ireland arevery reluctant to probe further in this area as survey non-response rates are liable to increase. In addition, it islikely that the age variable would be highly correlated with such a binary variable, e.g. younger respondentswould be much more likely to be living as married compared to older respondents.

Table 1Individual or household willingness to pay by version

Version

Response Individual first Household first Total

Household 296 (68.4%) 320 (72.2%) 616 (70.3%)Individual 137 (31.6%) 123 (27.8%) 260 (29.7%)

433 443 876


of the question subsequently claimed that their bid represented household WTP. Theresponses do not vary greatly between the two different versions of the survey.

In Version 1, where respondents are asked to give ‘‘your WTP’’, 68.4% of respondentssubsequently respond that this represents household WTP, whereas in Version 2 whererespondents are asked to give ‘‘your household WTP’’, 72.2% of respondents subsequentlyrespond that this represents household WTP. Therefore, 27.8% of respondents who wereasked to give ‘‘household WTP’’ instead give ‘‘individual WTP’’.

If attention is restricted to those (married or living as married) 338 respondents whowere willing to pay a positive amount of extra taxation in order to finance increases insocial welfare expenditures, a statistically significant difference emerges. In Version 1,approximately 76% of respondents provide household WTP, while in Version 2, approx-imately 84% of respondents provide household WTP. However, even for this sub-sampleof respondents, who are arguably more-focused on the detail of the questions as they pro-ceed to provide a positive WTP, approximately 16% of respondents who are asked to give‘‘household WTP’’ instead provide individual WTP. In summary, even CVM studies thatdo address agency ambiguity will yield at the least imprecise results.

The next objective is to examine the determinants of whether individuals respond asindividuals or as households, examining in particular the manner in which their householdfinances are structured. The respondents (876) who had previously stated that they weremarried or living as married were also asked about their household income-poolingarrangements; the results are displayed in Table 2.

In total 535 respondents answered the question; 8% of these respondents stated thatthey held completely separate finances, while 36.6% stated that they held completely inte-grated finances. Given rather small sub-samples sizes, a binary dummy variable was cre-ated that simply codes whether respondents held separate or joint accounts; 26.4% ofthe respondents were classified as holding separate accounts, while 73.6% of respondentswere classified as holding joint accounts.

Table 2Household financing structure

Frequency Valid (%) Cumulative (%)

Separate finances/rarely discuss 43 8.0 8.0Separate accounts 98 18.3 26.4Joint accounts 154 28.8 55.1Joint but other Money separate 44 8.2 63.3Joint accounts and conduct together 196 36.6 100

Total 535 100.0


Table 3 displays the marginal effects from two binary probit models, designed to exam-ine the determinants of whether a respondent states that his/her bid is equal to ‘‘individ-ual’’ (=0) or ‘‘household’’ (=1). The first model indicates that this decision is stronglyaffected by age and the presence of children, similar to the previous findings of Delaneyand O’Toole (2004) albeit in a different policy context. Thus, there is some degree of dif-ferential item functioning that can be explained in this literature. The marginal effects fromthe models give a strong indication of the scale of this difference in interpretation. Forexample, having children increases the likelihood that the respondent will respond withhousehold WTP by approximately 16%.

The inclusion of the binary measure of financial integration into the model demon-strates its significance; respondents from households where finances are conducted jointlyare approximately 18% more likely to respond with household WTP. Indeed, once thisvariable is included in the model, the effects of children and age are no longer statisticallysignificant. Interestingly, the negative effect of being male on responding as a householdbecomes statistically significant when one controls for household finances; this is similarto the findings of Delaney and O’Toole (2004), albeit in a different policy context, wheremen were more likely to respond as individuals than women. The key result, however,from this model is that respondents’ interpretation of CVM questions depends very signif-icantly and substantially on the manner in which household finances are structured.

4.2. Income pooling and preferences for child benefit

Within Table 4, columns 1–2 display the results of ordinary least squares regressions ofthe determinants of preferences for the level (in terms of the 1–7 scale) of (i) social welfare

Table 3Respondents answer as individuals (0) or households (1)?

dF/dx dF/dx

Age 0.017*** 0.0010.006 0.010

Age-squared 0.000 0.0000.000 0.000

Dependent children 0.156*** 0.0370.036 0.046

Joint finances – 0.184***– 0.054

Middle income 0.017 0.0270.039 0.043

High income �0.131** �0.0200.065 0.068

Household version 0.024 0.0610.035 0.041

Gender �0.04 �0.130***0.048 0.039

N 690 395Pseudo-R 0.070 0.090

Notes: Standard errors are given in parenthesis, *** indicates significance at the 1% level, ** indicates significanceat the 5% level, * indicates significance at the 10% level. The base category for income is low income. The basecategory for household version is individual version.

Table 4Determinants of preferences for government spending

Socialwelfare

Childbenefit

CBtaxable

CBprogressive

CBconditional

CBuniversal

Age 0.011 0.022* 0.012 �0.013 �0.022 �0.0350.015 0.013 0.021 0.022 0.022 0.022

Age-squared 0.000 0.000 0.000 0.000 0.000 0.0000.000 0.000 0.000 0.000 0.000 0.000

Dependentchildren

0.235** 0.427*** �0.471*** �0.437*** �0.282** 0.513***0.107 0.091 0.145 0.154 0.157 0.158

Male �0.035 �0.068 �0.087 �0.218** 0.143 �0.1290.103 0.087 0.140 0.148 0.151 0.152

Middle income �0.522*** �0.374*** �0.366** �0.522*** �0.699*** 0.523***0.109 0.092 0.148 0.158 0.161 0.162

High income �0.715*** �0.686*** �0.473 �1.439*** �1.451*** 1.176***0.230 0.195 0.308 0.328 0.334 0.336

Full-timeemployed

�0.406*** �0.205*** 0.284* 0.477 0.314 �0.0160.109 0.093 0.148 0.158 0.161 0.162

Male*high income �0.019 0.580** 0.486 1.472*** 1.151*** �1.253***0.311 0.264 0.420 0.447 0.456 0.458

Constant 4.998*** 4.759*** 2.852*** 5.164*** 5.260*** 4.554***0.302 0.255 0.410 0.433 0.441 0.444

N 827 827 827 827 827 827R2 0.090 0.070 0.040 0.060 0.070 0.050

Notes: Standard errors are given in parenthesis *** indicates significance at the 1% level, ** indicates significanceat the 5% level, * indicates significance at the 10% level. The base category for income is low income.


expenditure and (ii) child benefit expenditure. None of the conclusions are conditional onthe use of the ordinary least squares model and emerge from several other types of modelincluding ordered logit models. Columns 3–5 display the results of ordinary least squaresregressions of the determinants of preferences for three different schemes for making childbenefit more conditional on income. Column 6 displays the results of ordinary leastsquares regression of the determinants of preferences for the status quo with respect tochild benefit, i.e. leaving the child benefit as a universal (non-taxed) benefit generally paidto the mother.

The level of specific social welfare and child benefits payments are influenced in a fairlypredictable manner. For example, those with dependent children support an increase whilethose on middle and high incomes and those in full-time employment support a decrease.The gender variable does not significantly influence the level of either payment. However,males with higher incomes, compared to females with higher income, are statistically: (i)more in favour of making child benefit progressive in income (‘‘People with high incomes

should be given less child benefit than people with low incomes’’); (ii) more in favour of mak-ing child benefit conditional (‘‘Child benefit should only be paid to those who need it’’); and(iii), less in favour of maintaining the universality of child benefit (‘‘Child benefit should be

paid to everyone with children regardless of household income’’). There is no gender effect onthe proposal to tax child benefit (‘‘Child benefit should be part of taxable income’’). Thislatter result is consistent with the presence of obvious uncertainties with respect to howthis specific proposal would actually be implemented, e.g. the actual child benefit paymentto the parent (usually mother) could remain unchanged while the tax payment could bemade by the income earner (often male) within the household. The inclusion of the joint


finances binary variable does not impact on these results, apart of course from loweringthe sample size considerably.

5. Conclusions

There has been renewed interest in the use of the open-ended willingness to pay ques-tion as a method of eliciting respondents’ preferences for non-marketed goods (e.g. Ready,Navrud, & Dubourg, 2001). However, the use of the question should be accompanied by aserious caveat, namely whether the respondent interprets and models the implied valuationas being an individual or household valuation. In this paper, it is demonstrated thatrespondents, who are married or living as married, primarily but certainly not exclusivelymodel bids as being a household valuation. There was also a strong positive correlationbetween the level of household financial integration and the likelihood that a respondentmodels a WTP question as being a household decision. However, modelling respondents’WTP as being household valuations is problematic and may lead to a significant underes-timate of WTP for multi-person households. In particular, even when respondents wereasked for ‘‘household WTP’’, a substantial proportion of respondents interpreted this aseliciting individual valuation. Explicitly asking respondents to spend from the householdbudget constraint is no guarantee that they will do so, particularly if the respondent is notin a financially integrated relationship. The implication for CVM practice is that studiesthat assess WTP should include extra probes to inquire whether or not the respondentis giving household WTP.

In addition, it is demonstrated in this paper that women and men have different pref-erences for household public goods, lending further evidence to the view that the conceptof ‘‘household’’ preferences is problematic. Although in general women did not demon-strate significantly different patterns of support for specific government expenditures, therewas strong evidence that income and gender has an effect on determining preferences for aproposal to make child benefit a conditional payment on income. Introducing condition-ality would weaken the bargaining position of women in the top income bracket relative totheir partners and that for this reason women are more likely to oppose this than men.This is further and novel empirical evidence for how the economics of intra-householdbargaining determines the structure of preferences for public goods, and society-wide allo-cations. More specifically, it demonstrates that gender differences emerge significantly fora household public good when the provision of that good alters the intra-household enti-tlement to income between the partners.

Acknowledgements

We thank the two anonymous reviewers and the Associate Editor for their excellent andhelpful comments.

References

Alvarez, M., & McCaffrey, E. (2003). Are there sex differences in fiscal political preferences. Political Research

Quarterly, 56, 5–17.Burgoyne, C. (1995). Financial organisation and decision-making within Western ‘households’. Journal of

Economic Psychology, 16, 421–430.


Champ, P., Boyle, K., & Brown, C. (2003). A primer on nonmarket valuation. Dordrecht, The Netherlands:Kluwer Academic Publishers.

Delaney, L., & O’Toole, F. (2004). Irish public service broadcasting: A contingent valuation analysis. The

Economic and Social Review, 35, 321–350.Delaney, L., & O’Toole, F. (2006). Willingness to pay: Individual or household? Journal of Cultural Economics,

30, 305–309.Dupont, D. (2004). Do children matter? An examination of gender differences in environmental valuation.

Ecological Economics, 49, 273–286.Fong, C. (2001). Social preferences, self-interest, and the demand for redistribution. Journal of Public Economics,

82, 225–246.Hochman, H. M., & Rodgers, J. D. (1969). Pareto optimal redistribution. American Economic Review, 59,

542–557.Hoddinott, J., & Haddad, L. (1995). Does female income share influence household expenditure patterns? Oxford

Bulletin of Economics and Statistics, 57, 77–97.Kirchler, E., Rodler, C., Holzl, E., & Meier, K. (2001). Conflict and decision-making in close relationships: Love.

Money and daily routines. East Sussex: Psychology Press Ltd.Lampietti, J. (1999). Do husbands and wives make the same choices? Evidence from Northern Ethiopia.

Economics Letters, 62, 253–260.Mitchell, R., & Carson, R. (1989). Using surveys to value public goods: The contingent valuation method.

Washington, D.C: Resources for the Future Press.Pahl, J. (1995). His money, her money: Recent research on financial organisation in marriage. Journal of

Economic Psychology, 26, 361–376.Phipps, S. A., & Burton, P. A. (1992). What’s mine is yours? The influence of male and female incomes on

patterns of household expenditure. Economica, 65, 599–613.Quiggin, J. (1998). Individual and household willingness to pay for public goods. American Journal of Agricultural

Economics, 80, 58–63.Ready, R. C., Navrud, S., & Dubourg, W. R. (2001). How do respondents with uncertain willingness to pay

answer contingent valuation questions? Land Economics, 77, 315–326.Scheve, K., & Slaughter, M. (2006). Public opinion, international integration, and the welfare state. In P.

Bardhan, S. Bowles, & M. Wallerstein (Eds.), Globalization and egalitarian redistribution (pp. 217–260).Princeton: Princeton University Press.

Schultz, T. (1990). Testing the neoclassical model of family labor supply and fertility. Journal of Human

Resources, 25, 599–634.Strand, J. (2004). Public good valuation and intrafamily allocation. CESIFO Working Paper No. 1351 Category

8: resources and environment.Thomas, D. (1990). Intra-household resource allocation: An inferential approach. Journal of Human Resources,

25, 635–640.

Documents

Individual, household and gender preferences for social transfers